On topologically trivial automorphisms of compact Kähler manifolds and algebraic surfaces (Q2240517)

The article under review studies automorphism groups of compact Kähler surfaces up to various kinds of topological triviality, namely by means of smooth isotropy, homotopy as well as rational and integral cohomology. For \(X\) a compact Kähler manifold, the following sequence of gropus of holomorphic automorphisms can be considered: \[\mathrm{Aut}_0(X) \triangleleft\mathrm{Aut}_*(X)\triangleleft\mathrm{Aut}_{\#}(X)\triangleleft\mathrm{Aut}_{\mathbb{Z}}\triangleleft\mathrm{Aut}_{\mathbb{Q}}\triangleleft\mathrm{Aut}(X) \] Here, the rightmost group \(\mathrm{Aut}(X)\) is the full complex Lie group of holomorphic automorphisms of \(X\) and the leftmost sub-group \(\mathrm{Aut}_0(X)\) is its unit component. The latter is contained in the sub-group \(\mathrm{Aut}_*(X) = \mathrm{Aut}(X)\cap\mathrm{Diff}_0(X)\) of automorphisms which are \(\mathrm{C}^{\infty}\)-isotopic to the identity, which itself is a sub-group of the homotopically trivial automorphisms \(\mathrm{Aut}_{\#}(X)\). The two remaining groups in the sequence consist of the automorphisms acting trivially on the integral and rational cohomology of \(X\), respectively. Whilst the number of connected components of the full automorphism group, i.e., the index \([\mathrm{Aut}(X):\mathrm{Aut}_0(X)]\), may very well be infinite, it is known that for a compact Kähler manifold, the index \([\mathrm{Aut}_{\mathbb{Q}}(X):\mathrm{Aut}_0(X)]\) is finite. A synopsis of the main results of the article under review is that complex algebraic surfaces give rise to arbitrarily large indices \([\mathrm{Aut}_{\mathbb{Q}}(X):\mathrm{Aut}_{\mathbb{Z}}]\), \([\mathrm{Aut}_{\mathbb{Z}}(X):\mathrm{Aut}_{*}]\) and \([\mathrm{Aut}_{*}(X):\mathrm{Aut}_{0}(X)]\). Note that such surfaces must not be of general type (eventually), for the number of \(\mathbb{Q}\)-cohomologically trivial automorphisms of surfaces of general type is universally bounded. Although the homotopically trivial automorpihsms do play a role in intermediate arguments, the authors remark that the relation between \(\mathrm{Aut}_\#(X)\) and \(\mathrm{Aut}_*(X)\) ``is still mysterious'' and they pose the question whether \([\mathrm{Aut}_{\#}(X):\mathrm{Aut}_*(X)]\) is uniformly bounded. Section 2 of the article establishes six ``Principles'' which are the core of the arguments in the following chapters. The Principles 1, 2 and 3 are concerned with the way automorphisms interact with negative curves or fibrations and their special fibres. Principle 4 gives a detailed description of the \(\mathbb{Q}\)-cohomologically trivial automorphisms of surfaces isogenous to a product of (SIP) of unmixed type, i.e., surfaces of the form \(X = (C_1\times C_2)/\Delta_G\) where \(C_1,C_2\) are smooth curves of positive genus, the finite group \(G\) acts on both \(C_1\) and \(C_2\) and \(X\) is the quotient of the resulting diagonal action. Principle 5 establishes that along sequences of contractions of \((-1)\)-curves \(X = X_n \xrightarrow{f_n} X_{n-1} \cdots X_1 \xrightarrow{f_1} X_0\), the cohomologically trivial automorphisms of \(X\) descend to automorpisms of each \(X_k\) preserving the respective exceptional curves. This yields a description of \(\mathrm{Aut}_{\mathbb{Z}}(X)\) and \(\mathrm{Aut}_{\mathbb{Q}}(X)\) as sub-groups of \(\mathrm{Aut}_{\mathbb{Z}}(X_0)\) and \(\mathrm{Aut}_{\mathbb{Q}}(X_0)\), respectively. Sections 3, 4 and 5 of the article under review are dedicated to the constructions of series of surfaces with unbounded indices as mentioned above: The unboundedness of \([\mathrm{Aut}_{\mathbb{Q}}(X):\mathrm{Aut}_{\mathbb{Z}}(X)]\) is witnessed by an unmixed SIP of the form \(X=(B\times E)/\Delta_G\) where \(B\) is a certain hyperelliptic curve on which \(G = {\mathbb{Z}}/2{\mathbb{Z}}\) acts as the hyperelliptic involution and \(F\) is an elliptic curve on which \(G\) acts by translation along an element of order \(2\). Automorphisms cyclically permuting the ramification points of the hyperelliptic involution are shown to give rise to automorphisms of the same order on \(X\). The surfaces \(X\) witnessing the unboundedness of \([\mathrm{Aut}_{\mathbb{Z}}(X):\mathrm{Aut}_*(X)]\) are obtained as a blow-up of a suitably chosen point on a certain unmixed SIP \(Z = (C\times E)/\Delta_G\), where \(E\) is an elliptic curve, \(G\cong {\mathbb{Z}}/n{\mathbb{Z}}\) and the action of \(G\) on \(C\) has a fixed point. Note that the two series of examples proving the unboundedness of \([\mathrm{Aut}_{\mathbb{Q}}(X):\mathrm{Aut}_{\mathbb{Z}}(X)]\) and \([\mathrm{Aut}_{\mathbb{Z}}(X):\mathrm{Aut}_*(X)]\) are of Kodaira dimension \(1\). However, the surfaces constructed in Section 5, providing unbounded \([\mathrm{Aut}_*(X):\mathrm{Aut}_0(X)]\), are ruled surfaces and certain sequences of blow-ups of surfaces carrying a \(\mathbb C^*\)-action. The first construction is quite general and very concisely explained on p. 196. When applied to \(\mathbb P^2\), it yields an answer to questions raised by \textit{L. Meersseman} [Enseign. Math. (2) 63, No. 3--4, 263--272 (2017; Zbl 1435.32023)] and \textit{K. Oguiso} [Nagoya Math. J. 239, 110--122 (2020; Zbl 1440.14197)]. It should also be mentioned that the arguments presented refute a statement of \textit{C. A. M. Peters} [Journees de geometrie algebrique, Angers/France 1979, 249--267 (1980; Zbl 0444.14009)], cf. Remark 5.3 of the article under review. The second example is the minimal ruled surface \(X = \mathbb P(\mathcal O_E\oplus\mathcal O_E(D))\) where \(E\) is an elliptic curve and \(D\) is a divisor of even degree \(d=2m>0\). It is shown that \([\mathrm{Aut}_*(X):\mathrm{Aut}_0(X)]\geq m^2\) (cf. Theorem 5.4). A more in depth analysis of \(\mathbb P^1\)-bundles can be found in Section 7.1. Section 6 more deeply explores the relation between \(\mathrm{Aut}_{\mathbb{Q}}(X)\) and \(\mathrm{Aut}_*(X)\) for blow-ups of \(\mathbb P^2\), eventually proving that \(\mathrm{Aut}_{\mathbb{Q}}(X)=\mathrm{Aut}_*(X)\) for all rational surfaces. It is mentioned that the proof can be drastically simplified if the following interesting question can be answered positively: On a compact complex manifold, if a fixed point of a \(\mathrm{C}^\infty\)-isotopically trivial automorphism is blown up, does the lift of the automorphism along the blow-up also have the property of being \(\mathrm{C}^\infty\)-isotopically trivial? The concluding Section 7 contains an overview of the current knowledge about cohomologically trivial automorphisms of surfaces with Kodaira dimensions \(-\infty\), \(0\) and \(2\).